2021
DOI: 10.1016/j.jsv.2021.116196
|View full text |Cite
|
Sign up to set email alerts
|

Structural identification with physics-informed neural ordinary differential equations

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
45
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
4
2
2

Relationship

0
8

Authors

Journals

citations
Cited by 100 publications
(53 citation statements)
references
References 43 publications
(35 reference statements)
0
45
0
Order By: Relevance
“…Recent machine-learning methods have increasingly been influenced by physics to address issues with interpretability and prediction [ 45 , 46 ]. The proposed tools include sparse regression [ 47 , 48 ], neural networks [ 49 ], neural ordinary differential equations [ 50 ], simultaneous basis function approximation and parameter estimation [ 51 ] (see [ 45 ] for an extensive review). These physics-informed models are easier to generalize and also handle sparse and incomplete data better.…”
Section: Introductionmentioning
confidence: 99%
“…Recent machine-learning methods have increasingly been influenced by physics to address issues with interpretability and prediction [ 45 , 46 ]. The proposed tools include sparse regression [ 47 , 48 ], neural networks [ 49 ], neural ordinary differential equations [ 50 ], simultaneous basis function approximation and parameter estimation [ 51 ] (see [ 45 ] for an extensive review). These physics-informed models are easier to generalize and also handle sparse and incomplete data better.…”
Section: Introductionmentioning
confidence: 99%
“…ODEs can be used to simulate complex nonlinear systems which are difficult to model using simply physics-based models (Lai et al, 2021). A typical ODE system is written as…”
Section: Odementioning
confidence: 99%
“…A PINN approach is used by Lai et al (2021) for structural identification, using Neural Ordinary Differential Equations (Neural ODEs). Neural ODEs can be considered as a continuous representation of ResNets (Residual Networks), by using a neural network to parameterize a dynamical system in the form of ODE for an initial value problem (IVP):…”
Section: Odementioning
confidence: 99%
See 1 more Smart Citation
“…Besides that, a burgeoning discussion explores how to use PGML to unveil unknown governing equations and physics intuition based on data [293-295, 370, 423-425]. Recently, Lai et al [426] applied NODE to learn the governing structural dynamics and experimentally showed its effectiveness in a structure equipped with a negative stiffness device. Incipient research applied ECNN to learn the dynamics of the pendulum and multi-body problems [398,404,427,428].…”
Section: Physics Guided Machine Learningmentioning
confidence: 99%